Search

While recent text-to-image models can generate photorealistic images from text prompts that reflect detailed instructions, they still face significant challenges in accurately rendering words in the image. In this paper, we propose to retouch erroneous text renderings in the post-processing pipeline. Our approach, called Type-R, identifies typographical errors in the generated image, erases the erroneous text, regenerates text boxes for missing words, and finally corrects typos in the rendered words. Through extensive experiments, we show that Type-R, in combination with the latest text-to-image models such as Stable Diffusion or Flux, achieves the highest text rendering accuracy while maintaining image quality and also outperforms text-focused generation baselines in terms of balancing text accuracy and image quality.

We performed a search for the $K_L \to \pi^{0} \nu \bar{\nu}$ decay using the data taken in 2021 at the J-PARC KOTO experiment. With newly installed counters and new analysis method, the expected background was suppressed to $0.252\pm0.055_{\mathrm{stat}}$$^{+0.052}_{-0.067}$$_{\mathrm{syst}}$. With a single event sensitivity of $(9.33 \pm 0.06_{\rm stat} \pm 0.84_{\rm syst})\times 10^{-10}$, no events were observed in the signal region. An upper limit on the branching fraction for the decay was set to be $2.2\times10^{-9}$ at the 90% confidence level (C.L.), which improved the previous upper limit from KOTO by a factor of 1.4. With the same data, a search for $K_L \to \pi^{0} X^{0}$ was also performed, where $X^{0}$ is an invisible boson with a mass ranging from 1 MeV/$c^{2}$ to 260 MeV/$c^{2}$. For $X^{0}$ with a mass of 135 MeV/$c^{2}$, an upper limit on the branching fraction of $K_L \to \pi^{0} X^{0}$ was set to be $1.6\times10^{-9}$ at the 90% C.L., Comment: 6 pages, 4 figures; submitted for publication

We report the magnetic properties of CaCo$_2$TeO$_6$ as a Kitaev candidate. CaCo$_2$TeO$_6$ was synthesized through a topochemical process, wherein all Na$^+$ ions in Na$_2$Co$_2$TeO$_6$ were replaced with half the amount of Ca$^{2+}$ ions. This substitution brings the CoO$_6$ octahedra closer to an approximate cubic symmetry. CaCo$_2$TeO$_6$ exhibits antiferromagnetic ordering at $T_N \sim 13$ K, which is lower than $\sim$ 27 K observed for Na$_2$Co$_2$TeO$_6$. Notably, its magnetic order is suppressed in a relatively low magnetic field of approximately 4 T, indicating that non-Kitaev interactions can be restrained by reducing trigonal distortion. Our findings highlight the potential of CaCo$_2$TeO$_6$ as a viable platform for exploring Kitaev quantum spin liquids and pave the way for a deeper understanding of the fundamental mechanisms in Kitaev physics., Comment: 9 pages, 9 figures, accepted in Physical Review B

Recent advancements in foundation models show promising capability in graphic design generation. Several studies have started employing Large Multimodal Models (LMMs) to evaluate graphic designs, assuming that LMMs can properly assess their quality, but it is unclear if the evaluation is reliable. One way to evaluate the quality of graphic design is to assess whether the design adheres to fundamental graphic design principles, which are the designer's common practice. In this paper, we compare the behavior of GPT-based evaluation and heuristic evaluation based on design principles using human annotations collected from 60 subjects. Our experiments reveal that, while GPTs cannot distinguish small details, they have a reasonably good correlation with human annotation and exhibit a similar tendency to heuristic metrics based on design principles, suggesting that they are indeed capable of assessing the quality of graphic design. Our dataset is available at https://cyberagentailab.github.io/Graphic-design-evaluation ., Comment: Accepted to SIGGRAPH Asia 2024 (Technical Communications Track)

Aqueous solubility (AS) is a key physiochemical property that plays a crucial role in drug discovery and material design. We report a novel unified approach to predict and infer chemical compounds with the desired AS based on simple deterministic graph-theoretic descriptors, multiple linear regression (MLR) and mixed integer linear programming (MILP). Selected descriptors based on a forward stepwise procedure enabled the simplest regression model, MLR, to achieve significantly good prediction accuracy compared to the existing approaches, achieving the accuracy in the range [0.7191, 0.9377] for 29 diverse datasets. By simulating these descriptors and learning models as MILPs, we inferred mathematically exact and optimal compounds with the desired AS, prescribed structures, and up to 50 non-hydrogen atoms in a reasonable time range [6, 1204] seconds. These findings indicate a strong correlation between the simple graph-theoretic descriptors and the AS of compounds, potentially leading to a deeper understanding of their AS without relying on widely used complicated chemical descriptors and complex machine learning models that are computationally expensive, and therefore difficult to use for inference. An implementation of the proposed approach is available at https://github.com/ku-dml/mol-infer/tree/master/AqSol.

In this paper, we propose a novel family of descriptors of chemical graphs, named cycle-configuration (CC), that can be used in the standard "two-layered (2L) model" of mol-infer, a molecular inference framework based on mixed integer linear programming (MILP) and machine learning (ML). Proposed descriptors capture the notion of ortho/meta/para patterns that appear in aromatic rings, which has been impossible in the framework so far. Computational experiments show that, when the new descriptors are supplied, we can construct prediction functions of similar or better performance for all of the 27 tested chemical properties. We also provide an MILP formulation that asks for a chemical graph with desired properties under the 2L model with CC descriptors (2L+CC model). We show that a chemical graph with up to 50 non-hydrogen vertices can be inferred in a practical time.

In this paper, we first study what we call Superset-Subset-Disjoint (SSD) set system. Based on properties of SSD set system, we derive the following (I) to (IV): (I) For a nonnegative integer $k$ and a graph $G=(V,E)$ with $|V|\ge2$, let $X_1,X_2,\dots,X_q\subsetneq V$ denote all maximal proper subsets of $V$ that induce $k$-edge-connected subgraphs. Then at least one of (a) and (b) holds: (a) $\{X_1,X_2,\dots,X_q\}$ is a partition of $V$; and (b) $V\setminus X_1, V\setminus X_2,\dots,V\setminus X_q$ are pairwise disjoint. (II) For $k=1$ and a strongly-connected digraph $G$, whether $V$ is in (a) and/or (b) can be decided in $O(n+m)$ time and we can generate all such $X_1,X_2,\dots,X_q$ in $O(n+m+|X_1|+|X_2|+\dots+|X_q|)$ time, where $n=|V|$ and $m=|E|$. (III) For a digraph $G$, we can enumerate in linear delay all vertex subsets of $V$ that induce strongly-connected subgraphs. (IV) A digraph is Hamiltonian if there is a spanning subgraph that is strongly-connected and in the case (a)., Comment: 29 pages, 4 figures

We detail the optimized synthesis of the Mo$_3$O$_8$-type cluster Mott insulator (CMI) Na$_3$Sc$_2$(MoO$_4$)$_2$Mo$_3$O$_8$, which has been considered a candidate for realizing the spin liquid ground state. The optimized Na$_3$Sc$_2$(MoO$_4$)$_2$Mo$_3$O$_8$, characterized by x-ray diffraction, energy-dispersive x-ray spectroscopy, and magnetic and heat capacity measurements, exhibited an effective magnetic moment close to the ideal 1.73 $\mu_{\mathrm B}$ for $S = 1/2$ spin and magnetic ordering at $\sim$ 5 K. These observations categorize Na$_3$Sc$_2$(MoO$_4$)$_2$Mo$_3$O$_8$ as the second Mo$_3$O$_8$-type CMI to achieve a magnetic ground state, following Li$_2$InMo$_3$O$_8$. They highlight the stabilization of the magnetic ground state over the theoretically anticipated quantum spin liquid state through precise valence and chemical disorder tuning. Our findings challenge the existing theory that the magnetic ground state of Mo$_3$O$_8$-type CMIs is determined by the breathing parameter, instead showing that magnetic order is suppressed by spin defects. This study underscores the crucial role of chemical precision in investigating quantum magnetism. It suggests that precise tuning of valence states could induce magnetic ordering in previously nonmagnetic Mo$_3$O$_8$-type CMIs. Additionally, the negative findings regarding the existence of quantum spin liquids highlight the need for applied research and a reevaluation of our fundamental understanding of electronic states from both theoretical and experimental aspects., Comment: 8 pages, 5 figures, accepted in Physical Review Materials

Nb$_3$Cl$_8$, a cluster Mott insulator with a distinctive magnetic molecular orbital structure organized into a breathing kagome lattice, showcases critical phase transitions under specialized conditions. By transitioning from paramagnetic to nonmagnetic states below 90 K, we clarified this behavior through combined nuclear magnetic resonance and low-temperature X-ray diffraction studies, pointing to charge disproportionation as the driving force. Subsequent investigations via angle-resolved photoemission spectroscopy and first-principles calculations have disclosed topologically flat bands, confirming advanced electronic characteristics in Nb$_3$Cl$_8$. These discoveries not only deepen our comprehension of Mott insulators but also broaden our grasp of the dynamic interrelations among topology, electron interactions, and quantum phenomena in two-dimensional systems. The research on Nb$_3$Cl$_8$ thus lays foundational knowledge for advancing the exploration of quantum states in complex material systems, marking it as a critical model in the ongoing evolution of condensed matter physics., Comment: 16pages, 12figures

While single-cell RNA-seq enables the investigation of the celltype effect on the transcriptome, the pure tissue environmental effect has not been well investigated. The bias in the combination of tissue and celltype in the body made it difficult to evaluate the effect of pure tissue environment by omics data mining. It is important to prevent statistical confounding among discrete variables such as celltype, tissue, and other categorical variables when evaluating the effects of these variables. We propose a novel method to enumerate suitable analysis units of variables for estimating the effects of tissue environment by extending the maximal biclique enumeration problem for bipartite graphs to $k$-partite hypergraphs. We applied the proposed method to a large mouse single-cell transcriptome dataset of Tabala Muris Senis to evaluate pure tissue environmental effects on gene expression. Data Mining using the proposed method revealed pure tissue environment effects on gene expression and its age-related change among adipose sub-tissues. The method proposed in this study helps evaluations of the effects of discrete variables in exploratory data mining of large-scale genomics datasets.

By incorporating inert KCl into the Na$_2$IrO$_3$ + 2CuCl $\to$ Cu$_2$IrO$_3$ + 2NaCl topochemical reaction, we significantly reduced the synthesis temperature of Cu$_2$IrO$_3$ from the 350$^\circ$C reported in previous studies to 170$^\circ$C. This adjustment decreased the Cu/Ir antisite disorder concentration in Cu$_2$IrO$_3$ from $\sim$19$\%$ to $\sim$5$\%$. Furthermore, magnetic susceptibility measurements of the present Cu$_2$IrO$_3$ sample revealed a weak ferromagnetic-like anomaly with hysteresis at a magnetic transition temperature of $\sim$70 K. Our research indicates that the spin-disordered ground state reported in chemically disordered Cu$_2$IrO$_3$ is an extrinsic phenomenon, rather than an intrinsic one, underscoring the pivotal role of synthetic chemistry in understanding the application of Kitaev model to realistic materials., Comment: 7pages, 7 figures, 1 table, accepted in Journal of Physics: Condensed Matter

In this paper, we demonstrate a total disentanglement of font images. Total disentanglement is a neural network-based method for decomposing each font image nonlinearly and completely into its style and content (i.e., character class) features. It uses a simple but careful training procedure to extract the common style feature from all `A'-`Z' images in the same font and the common content feature from all `A' (or another class) images in different fonts. These disentangled features guarantee the reconstruction of the original font image. Various experiments have been conducted to understand the performance of total disentanglement. First, it is demonstrated that total disentanglement is achievable with very high accuracy; this is experimental proof of the long-standing open question, ``Does `A'-ness exist?'' Hofstadter (1985). Second, it is demonstrated that the disentangled features produced by total disentanglement apply to a variety of tasks, including font recognition, character recognition, and one-shot font image generation.

Deep Learning is advancing medical imaging Research and Development (R&D), leading to the frequent clinical use of Artificial Intelligence/Machine Learning (AI/ML)-based medical devices. However, to advance AI R&D, two challenges arise: 1) significant data imbalance, with most data from Europe/America and under 10% from Asia, despite its 60% global population share; and 2) hefty time and investment needed to curate proprietary datasets for commercial use. In response, we established the first commercial medical imaging platform, encompassing steps like: 1) data collection, 2) data selection, 3) annotation, and 4) pre-processing. Moreover, we focus on harnessing under-represented data from Japan and broader Asia, including Computed Tomography, Magnetic Resonance Imaging, and Whole Slide Imaging scans. Using the collected data, we are preparing/providing ready-to-use datasets for medical AI R&D by 1) offering these datasets to AI firms, biopharma, and medical device makers and 2) using them as training/test data to develop tailored AI solutions for such entities. We also aim to merge Blockchain for data security and plan to synthesize rare disease data via generative AI. DataHub Website: https://medical-datahub.ai/, Comment: 5 pages, 3 figures, accepted to SPIE Medical Imaging 2024

The purpose of this paper is to enable the conversion between machine-printed character images (i.e., font images) and handwritten character images through machine learning. For this purpose, we propose a novel unpaired image-to-image domain conversion method, CycleDM, which incorporates the concept of CycleGAN into the diffusion model. Specifically, CycleDM has two internal conversion models that bridge the denoising processes of two image domains. These conversion models are efficiently trained without explicit correspondence between the domains. By applying machine-printed and handwritten character images to the two modalities, CycleDM realizes the conversion between them. Our experiments for evaluating the converted images quantitatively and qualitatively found that ours performs better than other comparable approaches.

This study analyzes the relationship between non-verbal information (e.g., genres) and text design (e.g., font style, character color, etc.) through the classification of book genres using text design on book covers. Text images have both semantic information about the word itself and other information (non-semantic information or visual design), such as font style, character color, etc. When we read a word printed on some materials, we receive impressions or other information from both the word itself and the visual design. Basically, we can understand verbal information only from semantic information, i.e., the words themselves; however, we can consider that text design is helpful for understanding other additional information (i.e., non-verbal information), such as impressions, genre, etc. To investigate the effect of text design, we analyze text design using words printed on book covers and their genres in two scenarios. First, we attempted to understand the importance of visual design for determining the genre (i.e., non-verbal information) of books by analyzing the differences in the relationship between semantic information/visual design and genres. In the experiment, we found that semantic information is sufficient to determine the genre; however, text design is helpful in adding more discriminative features for book genres. Second, we investigated the effect of each text design on book genres. As a result, we found that each text design characterizes some book genres. For example, font style is useful to add more discriminative features for genres of ``Mystery, Thriller \& Suspense'' and ``Christian books \& Bibles.''

Fonts convey different impressions to readers. These impressions often come from the font shapes. However, the correlation between fonts and their impression is weak and unstable because impressions are subjective. To capture such weak and unstable cross-modal correlation between font shapes and their impressions, we propose Impression-CLIP, which is a novel machine-learning model based on CLIP (Contrastive Language-Image Pre-training). By using the CLIP-based model, font image features and their impression features are pulled closer, and font image features and unrelated impression features are pushed apart. This procedure realizes co-embedding between font image and their impressions. In our experiment, we perform cross-modal retrieval between fonts and impressions through co-embedding. The results indicate that Impression-CLIP achieves better retrieval accuracy than the state-of-the-art method. Additionally, our model shows the robustness to noise and missing tags.

This paper addresses the challenging task of estimating font impressions from real font images. We use a font dataset with annotation about font impressions and a convolutional neural network (CNN) framework for this task. However, impressions attached to individual fonts are often missing and noisy because of the subjective characteristic of font impression annotation. To realize stable impression estimation even with such a dataset, we propose an exemplar-based impression estimation approach, which relies on a strategy of ensembling impressions of exemplar fonts that are similar to the input image. In addition, we train CNN with synthetic font images that mimic scanned word images so that CNN estimates impressions of font images in the wild. We evaluate the basic performance of the proposed estimation method quantitatively and qualitatively. Then, we conduct a correlation analysis between book genres and font impressions on real book cover images; it is important to note that this analysis is only possible with our impression estimation method. The analysis reveals various trends in the correlation between them - this fact supports a hypothesis that book cover designers carefully choose a font for a book cover considering the impression given by the font.

Recent diffusion-based generative models show promise in their ability to generate text images, but limitations in specifying the styles of the generated texts render them insufficient in the realm of typographic design. This paper proposes a typographic text generation system to add and modify text on typographic designs while specifying font styles, colors, and text effects. The proposed system is a novel combination of two off-the-shelf methods for diffusion models, ControlNet and Blended Latent Diffusion. The former functions to generate text images under the guidance of edge conditions specifying stroke contours. The latter blends latent noise in Latent Diffusion Models (LDM) to add typographic text naturally onto an existing background. We first show that given appropriate text edges, ControlNet can generate texts in specified fonts while incorporating effects described by prompts. We further introduce text edge manipulation as an intuitive and customizable way to produce texts with complex effects such as ``shadows'' and ``reflections''. Finally, with the proposed system, we successfully add and modify texts on a predefined background while preserving its overall coherence.

Fonts have huge variations in their styles and give readers different impressions. Therefore, generating new fonts is worthy of giving new impressions to readers. In this paper, we employ diffusion models to generate new font styles by interpolating a pair of reference fonts with different styles. More specifically, we propose three different interpolation approaches, image-blending, condition-blending, and noise-blending, with the diffusion models. We perform qualitative and quantitative experimental analyses to understand the style generation ability of the three approaches. According to experimental results, three proposed approaches can generate not only expected font styles but also somewhat serendipitous font styles. We also compare the approaches with a state-of-the-art style-conditional Latin-font generative network model to confirm the validity of using the diffusion models for the style interpolation task.

Swarm intelligence algorithms have traditionally been designed for continuous optimization problems, and these algorithms have been modified and extended for application to discrete optimization problems. Notably, their application in feature selection for machine learning has demonstrated improvements in model accuracy, reduction of unnecessary data, and decreased computational time. This study proposes the Frog-Snake prey-predation Relationship Optimization (FSRO) algorithm, inspired by the prey-predation relationship between frogs and snakes for application to discrete optimization problems. The algorithm models three stages of a snake's foraging behavior "search", "approach", and "capture" as well as the frog's characteristic behavior of staying still to attract and then escaping. Furthermore, the introduction of the concept of evolutionary game theory enables dynamic control of the search process. The proposed algorithm conducts computational experiments on feature selection using 26 types of machine learning datasets to analyze its performance and identify improvements. In computer experiments, the proposed algorithm showed better performance than the comparison algorithms in terms of the best and standard deviation of fitness value and Accuracy. It was also proved that dynamic search control by evolutionary game theory is an effective method, and the proposed algorithm has the ability of a well-balanced search, achieving the two objectives of improving accuracy and reducing data.

Plankton recognition provides novel possibilities to study various environmental aspects and an interesting real-world context to develop domain adaptation (DA) methods. Different imaging instruments cause domain shift between datasets hampering the development of general plankton recognition methods. A promising remedy for this is DA allowing to adapt a model trained on one instrument to other instruments. In this paper, we present a new DA dataset called DAPlankton which consists of phytoplankton images obtained with different instruments. Phytoplankton provides a challenging DA problem due to the fine-grained nature of the task and high class imbalance in real-world datasets. DAPlankton consists of two subsets. DAPlankton_LAB contains images of cultured phytoplankton providing a balanced dataset with minimal label uncertainty. DAPlankton_SEA consists of images collected from the Baltic Sea providing challenging real-world data with large intra-class variance and class imbalance. We further present a benchmark comparison of three widely used DA methods.

We have successfully synthesized four Mo$_3$O$_8$-type cluster Mott insulators (CMI) by intercalating lithium into nonmagnetic precursors to regulate the Mo$_3$ cluster valence. The resulting materials are Li$_{1+x}$$R$Mo$_3$O$_8$ ($R$ = Sc, Y, Lu) and Li$_x$Zn$_2$Mo$_3$O$_8$. Our magnetic susceptibility measurements revealed that these materials display characteristics akin to a valence bond glass state and suggest the presence of short-range ordering when the Mo$_3$ cluster valence approximates its ideal value. These findings challenge the prevailing belief that the plaquette charge ordering state is an inherent feature of Mo$_3$O$_8$-type CMI. Instead, they underscore the importance of Mo$_3$ cluster valence in determining the physical properties of these systems. These insights furnish a fresh understanding of the Mo$_3$O$_8$-type CMI and open new research opportunities in highly frustrated magnetism., Comment: 13page, 11figures, manuscript accepted for publication in Physical Review B

Shortcut reasoning is an irrational process of inference, which degrades the robustness of an NLP model. While a number of previous work has tackled the identification of shortcut reasoning, there are still two major limitations: (i) a method for quantifying the severity of the discovered shortcut reasoning is not provided; (ii) certain types of shortcut reasoning may be missed. To address these issues, we propose a novel method for identifying shortcut reasoning. The proposed method quantifies the severity of the shortcut reasoning by leveraging out-of-distribution data and does not make any assumptions about the type of tokens triggering the shortcut reasoning. Our experiments on Natural Language Inference and Sentiment Analysis demonstrate that our framework successfully discovers known and unknown shortcut reasoning in the previous work.

Lip segmentation is crucial in computer vision, especially for lip reading. Despite extensive face segmentation research, lip segmentation has received limited attention. The aim of this study is to compare state-of-the-art lip segmentation models using a standardized setting and a publicly available dataset. Five techniques, namely EHANet, Mask2Former, BiSeNet V2, PIDNet, and STDC1, are qualitatively selected based on their reported performance, inference time, code availability, recency, and popularity. The CelebAMask-HQ dataset, comprising manually annotated face images, is used to fairly assess the lip segmentation performance of the selected models. Inference experiments are conducted on a Raspberry Pi4 to emulate limited computational resources. The results show that Mask2Former and EHANet have the best performances in terms of mIoU score. BiSeNet V2 demonstrate competitive performance, while PIDNet excels in recall but has lower precision. Most models present inference time ranging from 1000 to around 3000 milliseconds on a Raspberry Pi4, with PIDNet having the lowest mean inference time. This study provides a comprehensive evaluation of lip segmentation models, highlighting their performance and inference times. The findings contribute to the development of lightweight techniques and establish benchmarks for future advances in lip segmentation, especially in IoT and edge computing scenarios.

The sample quality of the hyperhoneycomb lattice Kitaev magnet $\beta$-ZnIrO$_3$ was successfully improved by removing the maximum amount of moisture from the reaction ampoule. The X-ray diffraction structural analysis of the high-quality sample confirmed the presence of $P$2$_1$/$c$ superlattice peaks of the Fddd structure in the original structural model. These peaks could not be distinguished due to the presence of impurities in the low-quality sample in a previous study. The structural analysis based on this monoclinic crystal structure model showed no chemical disorder, suggesting that the observed spin liquid type behavior is an intrinsic property unrelated to bond randomness. The details of the $\beta$-ZnIrO$_3$ structure revealed in this study will stimulate the further investigation of Kitaev physics., Comment: 4pages, 3figures

When we compare fonts, we often pay attention to styles of local parts, such as serifs and curvatures. This paper proposes an attention mechanism to find important local parts. The local parts with larger attention are then considered important. The proposed mechanism can be trained in a quasi-self-supervised manner that requires no manual annotation other than knowing that a set of character images is from the same font, such as Helvetica. After confirming that the trained attention mechanism can find style-relevant local parts, we utilize the resulting attention for local style-aware font generation. Specifically, we design a new reconstruction loss function to put more weight on the local parts with larger attention for generating character images with more accurate style realization. This loss function has the merit of applicability to various font generation models. Our experimental results show that the proposed loss function improves the quality of generated character images by several few-shot font generation models., Comment: Accepted at ICDAR WML 2023